Covering robots entirely with electronic skins (e-skins) has been a long-term aspiration. However, the number and density of the tactile sensors are limited by factors such as routing complexity, signal latency, and power consumption. Inspired by neural signal processing, we present a multichannel capacitive interface circuits that read out the sparse tactile signals in an event-driven (ED) manner. A self-reconfigurable successive-approximation-register capacitance-to-digital converter (SR SAR CDC) is proposed to reduce power consumption in passive-sampling (PS) mode and to achieve high resolution in noise-shaping (NS) mode. To support parallel recording in NS data converters, we propose a 2 N^{N} oversampling-rate (2 N^{N} -OSR) code division multiplexing (CDM) technique. Besides, an ED inter-integrated circuit (ED-I2C) protocol is proposed to reduce communication data burden and routing complexity. Fabricated in 55-nm CMOS technology, the multichannel capacitive sensor interface occupies a chip area of 2.244 mm 2^{\textrm {2}} . At a frame rate of 3.47 kHz, this work achieves a measured single-channel signal-to-noise-ratio (SNR) of 51 dB and a power consumption of 7.2 μW\mu \text{W} /Ch in NS mode. The stand-by power can be further reduced to 1.6 μW\mu \text{W} /Ch in PS mode. The ED sensing protocol significantly reduces the data burden for processing and communication, supporting a high-density integration of 4.8 sensors/cm 2^{\textrm {2}} and 16 sensors/wire.

中文翻译：

---

基于自重构 SAR 电容数字转换的全动态事件驱动电容传感器接口电路，用于高密度机器人触觉传感


用电子皮肤（e-skin）完全覆盖机器人一直是一个长期的愿望。然而，触觉传感器的数量和密度受到路由复杂性、信号延迟和功耗等因素的限制。受神经信号处理的启发，我们提出了一种多通道电容接口电路，它以事件驱动（ED）的方式读出稀疏的触觉信号。提出了一种自重构逐次逼近寄存器电容数字转换器（SR SAR CDC），以降低无源采样（PS）模式下的功耗并在噪声整形（NS）模式下实现高分辨率。为了支持 NS 数据转换器中的并行记录，我们提出了 2 N^{N} 过采样率 (2 N^{N} -OSR) 码分复用 (CDM) 技术。此外，还提出了ED内部集成电路（ED-I2C）协议来减少通信数据负担和路由复杂性。多通道电容传感器接口采用 55 nm CMOS 技术制造，占用芯片面积 2.244 mm 2^{\textrm {2}} 。在3.47 kHz的帧速率下，该工作在NS模式下实现了51 dB的实测单通道信噪比（SNR）和7.2 μW\mu \text{W} /Ch的功耗。 PS模式下待机功耗可进一步降低至1.6 μW\mu \text{W} /Ch。 ED传感协议显着减轻了处理和通信的数据负担，支持4.8个传感器/cm 2^{\textrm {2}}和16个传感器/线的高密度集成。